Defining the Abatement Cost in Defining the Abatement Cost in Presence of Learning-by-doing: Presence of Learning-by-doing: Application to the Fuel Cell Application to the Fuel Cell Electric Vehicle Electric Vehicle Anna Creti Anna Creti (Université Paris Dauphine and Ecole (Université Paris Dauphine and Ecole Polytechnique) Polytechnique) Energy Energy and Climate and Climate Joint work with Conference Confer ce A. Kotelnikova (E. Polytechnique), G. Meunier (INRA), J-P Ponssard (E. Polytechnique and Toulouse, Sept Toulouse, Sept CNRS) 2015 2015 1
EUROPEAN ROAD MAPS FOR THE EUROPEAN ROAD MAPS FOR THE European roadmaps for FCEV DEPLOYMENT OF FCEV DEPLOYMENT OF FCEV
FCEV may reach a substantial market share by 2050 iff iff - Manufacturing cost decreases (Toyota Mirai sells at 65 k€ in 2015) - Clean and cheap H2 production (renewables + electrolysis) - Network for H2 distribution is deployed Some references: Mc Kinsey (2010) Bruegel (2012), Rösler et al. (2014), Fuelling Europe’s Future (2014) 3
Marginal abatement cost (MAC) curves are a standard tool in environmental economics Practical assessment of a MAC in a dynamic setting is not straightforward We contribute to the debate on the MAC curves by extending of the standard concept of static abatement costs to a dynamic one. ◦ To do so, we introduce learning-by-doing learning-by-doing together with cost cost convexity convexity, as these two characteristics adequately describe many low-carbon technologies such as renewables
MC Basic MAC Basic MAC p CO2 t p CO2 t* = (c o – c n )/e New p CO2 t Old t* t q* n With With convexit convexity a range range of static of static MAC MAC for a traject for a trajectory ry p CO2 New t t 2 * t 1 * Old t 1 * t 2 * t q* n
Convexity induces a transition in the deployment 1. Along the optimal trajectory the static MACs are 2. equal to the cost of carbon With learning-by-doing a learning effect has to be 3. introduced in this equality Does the transition starts earlier or later? 4. Suppose the trajectory is given, when to start the 5. transition? What if there is more than one sector? 6. In this paper we explore (revisit) questions 1 to 5
Our modeling choices are close to papers analysing the role of cost convexities in the dynamics of abatement costs: ◦ The shape and the structure of MAC curves are sensitive to many factors, in particular to technical change (Goulder and Mathai, 2000; Manne and Richels, 2004) ◦ Amigues et al. (2014) analyze the optimal timing of carbon capture and storage policies under increasing returns to scale and find that the carbon capture of the emissions should start earlier than under a constant average cost assumption. ◦ Bramoullé and Olson (2005) examine why infant technologies may be preferred to mature technologies because of learning-by-doing and cost convexity ◦ Vogt-Schilb et al. (2012) introduce convexities in the cost functions of various sectors and show that the date at which the respective renewable technologies should be launched depends on the degree of the cost convexities.
Our contribution is complementary to large scale bottom-up models which have integrated endogenous technological change with learning-by-doing (MESSAGE, MARKAL and POLES) or sectorial ones (Rösler et al., 2014). Our approach allows to analytically characterise optimal deployment trajectories, and to calibrate them in the context of an empirical case study: ◦ the transition from Internal Combustion Engine (ICE) vehicles to Fuel Cell Electric Vehicle (FCEV).
We analyse the transition issue as the whole deployment phase of the new technology in substitution to an old polluting technology. The optimal optimal trajectory rajectory is a smooth transition in which green cars progressively replace old cars. ◦ During During the the transition ransition the CO2 price should be equal to the sum of two terms: the difference between the cost of the marginal green car and a polluting car; and the learning benefits over the future. ◦ At At the the end nd of of the the transi ransition the fleet is completely green. We characterize the second best MAC by addressing the following questions: ◦ At which date the date the new technology should be launched? At which rate would its deployment occurs? As for the FCEV case study, the dynamic abatement cost which allows to launch hydrogen car deployment in 2015 is 53 53 €/tCO2. €/tCO2.
The model Optimal abatement trajectories Launching date and deployment strategies: second best Illustration: the FCEV case Further research
with
To analyse how the sub-optimality of deployment scenario impacts the launching date, we describe an optimal trajectory as an optimal deployment scenario and the associated optimal launching date.
Questions When to launch the deployment of the program as calibrated from industry data Why the static abatment cost is a poor indicator What is the appropriate abatement cost What if to launch the program in 2015 18
Fig 1: TCO in €/km per year Fig 1: TCO in €/km per year FCEV car park in million units 0,8 15% market share in 2050 Tax ICE or 0,6 Infrastructure FCEV 0,4 8 fuel cost 7 6 0,2 5 4 3 0 2 1 0 2020 2020 2030 2030 2050 2050 Manufacturing 2015 2020 2025 2030 2035 2040 2045 2050 2055 2060 ICE FCEV ICE FCEV ICE FCEV From an exogenous ramp up scenario Fig 2: Fig 2: TCO TCO authors versus Rösler et authors versus Rösler et al. al. (2014) (2014) to a simple (Excel) model for 0,8 - manufacturing, fuel, 0,6 infrastructure costs 0,4 Rösler et al. - CO2 emissions 0,2 authors - calibrated on previous studies 0 2020 ICE 2020 2040 ICE 2040 FCEV FCEV TCO=Total Cost of Ownership
Fig 6: Static abatement cost (€/t) 2000 1500 1000 €/t 500 0 ‐ 500 The SMAC at year n depends on the earlier deployment How to take care of this inconsistency? 20
Methodology Take the deployment as a « green plant » to be launched in 2050 Assume infinite life duration • and no further cost (TCOs converge) Compute emissions avoided in • 2050 Calculation (no TICPE) The optimal timing is 53 €/t = Capital cost* 4% / 2030 13,2 Mt emissions avoided in 2050 TICPE= gasoline tax 21
4 parameter unit Base case Table 2 Target analysis target Discounted cost for the scenario up to 2050 M€ 17 511 10 582 Avoided CO2 emissions in 2050 Mt/year 13,2 14,1 Dynamic abatement cost €/t 53 30 Market size in % of total car park % 15% 20% Gasoline price (yearly rate of increase) % 1,4% 1,8% Manufacturing cost (FCEV vs ICE in 2050) % 11,3% 9,8% Hydrogen production cost in 2050 €/kg 6,8 6,2 Base Suggested case Target 22
The static abatement cost is a poor instrument for policy analysis; it decreases from 1600 €/tCO2 in 2020 to 650 €/tCO2 in 2030, is null in 2042 and then becomes negative Our methodology integrates learning-by-doing, provides a simple summary proxy for policy analysis and delivers an attractive framework for simulations ◦ The dynamic abatement cost for the reference scenario is 53 €/tCO2 in 2015 ◦ Assume the normative social cost of carbon is 30 €/tCO2 in 2015 (Quinet 2009, Quinet 2013) The optimal launching date should be postponed from 2015 to 2030 Or some key parameters of the scenario should be strenghtened Limitations and Extensions ◦ Financing ; complementary innovations 23
anna.creti@dauphine.fr
Recommend
More recommend
